The neural network has determined the gait characteristics of animals with Parkinson’s disease.
Bioinformatics from Osaka University under the leadership of Takui Maekawa have developed a neural network that detects Parkinson’s disease by movement. The authors note that the characteristics of the disorder persist among species in evolution, therefore, it is possible to study the manifestations of the disease on model organisms.
The new model defines the features of movement for humans, mice, beetles, and roundworms. According to the device, this is a domain-adversarial neural network: it receives data about the parameters of movement, in particular, about speed and trajectory, and at the output it gives data about whether this corresponds to the norm, as well as about who has or is not found to have a disorder.
As a result, it turned out that despite completely different locomotion systems, the mouse and the worm with Parkinson’s disease were unable to maintain a high speed for a long time. Later, similar information was confirmed for humans. The worm and the human both moved unstably at high acceleration. And the worm and the beetle could not make smooth turns.
Such patterns could be used to further investigate dopamine deficiency, and the neural network could be applied to other motor disorders as well.